Bad Tearing in 3D Mark 11 - water cooled / crossfired HD 6990s

R

roadczar

Limp Gawd
Joined
Mar 6, 2011
Messages
131
Edit: changed tearing to artifacts -

Here is my build in progress: http://th10.roadczar.com/

After putting the system together having a pair of water cooled and crossfired HD 6990s.
I’m getting horrible artifacts in 3D Mark 11. This did not happen with same cards in my previous air cooled setup.

Looks like I erroneously mounted the second HD 6990 in an x8 PCI-E slot instead of the 16x slot. Another possibility is poor vram contacts to the water block. The GPU temps are OK, <50c under full load, not sure about the vram. Next mod is to go to the 3 way bridge and reseat the water block on the second card with fresh thermal tape.

Do you guys think that placing second HD 6990 in an 8x PCI-E slot instead of 16x slot can cause this tearing or can this be a vram temperature issue?

I already eliminated drivers from the equation and narrowed it down to one of the cards by using PCI-e switching option of Rampage IV.

Thanks .

SAM_0506-1024x768.jpg
 
Last edited:
Hmm dude, tearing has to do with vsync.

Maybe you are confusing tearing for something else?
 
Yea tearing is a vsync issue. 8 vs 16 PCIe lanes has nearly no impact on performance at all so that couldn't be the source of any issues either.
 
http://youtu.be/uN7vlQdkNNQ

Here is the vid, you tell me. Like I said in my first post, with my prior x58 setup the cards ran without issues. Now on x79 I get this. What else can this be?
I'd rather not have to tear up the loop to switch slots.
 
May want to get the latest bios and chipset drivers. Does it only do it in eyefinity? Or will it do it on a single display?
 
that's artifacts and fps lag, looks like something is broken or just not enough juice for those settings. + ur card is clocked too high or overheating?
 
you can get GPU-Z see the vram temps also. As stated before thats not tearing thats artifacts and I would look at the vrm temps. What slot its in should make no difference.
 
Last edited:
One more thing -

I can run MSI Kombustor on extreme setting with Prime95 running 12 threads.
Even running it for hours there are no artifacts and the temps never reach over 50c on either the CPU or any of GPUs.

I will try re-seating the GPU waterblocks this weekend.

I’m stomped. :confused:

Another video:
http://youtu.be/l8C8CfHLfpU
 
roadczar said:
I can run MSI Kombustor on extreme setting with Prime95 running 12 threads.
Click to expand...
At the same time? That's just going to create a massive CPU bottleneck in Kombustor and take a lot of the load off of your GPUs.

roadczar said:
I already eliminated drivers from the equation and narrowed it down to one of the cards by using PCI-e switching option of Rampage IV.
Click to expand...
If you can confirm that one of the cards alone artifacts and the other does not, then you can pretty much rule out any kind of software issue.
 
Did you have them watercooled in your X58 setup?

EDIT: I see they were air cooled previously. I would look at the waterblock, make sure the VRMs are making good contact.
 
I would double check the heatsinks on the RAM for the cards.
 
Thats not tearing, thats artifacting, you have a heat problem somewhere...
Probably ram, either too much heat or its going bad.
 
I doubt its the ram overheating. On the vrms did you put a thermal pad on it. This is what they look like.

radeon_hd6990_gpu_vrm.jpg
 
Thanks guys.

I am leaning to a vram issue as well. If I get all the parts in time then the plan is to drain the loop, re-seat the block and move the card to a 16x slot. I have to do this in one shot because draining and refiling this beast is not a trivial affair which is caused by the top mounted radiators. I'm reconsidering going to a bottom radiator configuration.

I will also be adding 3 temperature sensors and 2 flow sensors to the loop which should make debugging temp related issues a little easier.

The flow sensors will be added at the pump exit and at the reservoir entrance which should give me a nice delta readings.

The temp sensors will be added at the pump exit, at the CPU exit, and at the last GPU exit this should be interesting as well.
 
If vrams are the cause would I not see the tearing/artifacting in MSI Kombustor as well?
 
roadczar said:
If vrams are the cause would I not see the tearing/artifacting in MSI Kombustor as well?
Click to expand...

Not necessarily. I had this issue with my SLI 8800GTXs years ago, and I would get the effect sometimes while running the 3d benchmarks and while playing Crysis, but not necessarily with other games. It also happened most of the time after OCing. But I replaced the heatsinks and used actual thermalpaste rather than the cheap tape the heatsinks came with and it all worked fine after that. It is just one possible issue though, but its the most common I have heard of from those doing WC on their gfx cards.
 
Wait a few days, after the 7000 series comes out and RMA the cards. Maybe if you're lucky they'll send you 7970s as replacements lol.
 
Simple test, underclock your Vram below default as low as you can and run the benchy again. It will heat much slower, grab a stop watch and start it at your settings now, if the artifacting is delayed after you underclock the Vram then it is an overheating issue.
 
Also have you tried putting the air cooler back on and testing it? Then you know its a cooling issue and not defective hardware.
 
I've tried to put after market air coolers on for friends in the past or have watched a few guys put these on and the results have always been horrible. Instant artifacting brought on because of damage, somehow, to the video card. I will never do that again. I only hope this guy hasn't ruined his card(s).
 
Shadowmonk said:
Simple test, underclock your Vram below default as low as you can and run the benchy again. It will heat much slower, grab a stop watch and start it at your settings now, if the artifacting is delayed after you underclock the Vram then it is an overheating issue.
Click to expand...

I did try it with same results. The artifacting is instant even after a cold boot. Thanks.
 
qbanb8582 said:
Also have you tried putting the air cooler back on and testing it? Then you know its a cooling issue and not defective hardware.
Click to expand...

I will try it as the last resort. I'm still hoping this might be BIOS teething issue.
 
After watching the video, and noticing this is artifacting, I believe you have one of 2 scenarios:

1) Overclocked the GPU Mem to high

or

2) Faulty/broken connection to one or more memory chips on the GFX card - can happen when you overdo the screws when fitting the waterblock.
 
It really looks like artifacting, however if you can use kombustor with no problems that makes me wonder. Have you played any demanding games on this system yet? If nothing else is having issues, I would have to say its a problem with 3D Mark 11.
 
MrBean_Oz said:
After watching the video, and noticing this is artifacting, I believe you have one of 2 scenarios:

1) Overclocked the GPU Mem to high

or

2) Faulty/broken connection to one or more memory chips on the GFX card - can happen when you overdo the screws when fitting the waterblock.
Click to expand...

1 At the moment the cards are not overclocked
2 That would not be good. I will inspect during the tear down.
 
Direfox said:
It really looks like artifacting, however if you can use kombustor with no problems that makes me wonder. Have you played any demanding games on this system yet? If nothing else is having issues, I would have to say its a problem with 3D Mark 11.
Click to expand...

This was an interesting test. Not too many games, but:
BF3 would not launch - I am sure it is a separate issue
BFBC2 - runs fine, but with bad flickering. Not sure if related since I have seen it before caused by driver issues.
BioShock - no issues.
 
Bioshock has no issues simple due to the fact it does not use the "suspected" damaged memory chip(s) would be my guess. In fact, I'm willing to bet you will have very little issues with older games since they will use 512 to no more than 1gig of video frame memory. The newer the game, the more issues. I think you've figured out your problem.
 
I'm almost done re-configuring my loop and should be able to test in couple of hours.
 
You've had lots of advice in this thread, but this thread reminded me of a problem Kyle had awhile back where he had over tightened the heat-sink screws on something. He loosened them off some and his problem went away. Might be something to consider.
 
I'm gonna go with a bad block install, overtightened screws more than likely, bad contact is a possibility too. Try each card one at a time (I know, pain in the ass in a loop)

You can narrow the problem down to a specific card then go back to air cooling on the problematic card. If the problem goes away reapply the cooler being extra careful and make sure you have 100% contact and not to overtighten (or undertighten) )anything
 
I would say the first thing you should do is unplug a card and try a single card, and see which card is bad so you can RMA it.
 
So… I’ve spent couple of days reinstalling water block, rebuilding the loop, moving the card to the 16x slot, etc. Not everything was related to debugging the artifacts as I wanted to reconfigure some of the items regardless.

End result? You guessed it – slightly better, but still artifacts in the second card.

The neat thing about the Rampage IV is it has some dip switches that allows you to isolate different PCI-E slots. These were put to good use. I spent few hours messing around with various MSI settings and running 3D Mark 11 benchmarks. I kept on thinking the card was not getting enough voltage. Since I have more than enough cooling capacity I decided to max it out – the artifacts got worse.

After reducing voltage from 1175mV (stock with OC BIOS) to 1120mV the artifacts were completely gone!

Looks like my primary card likes 1175mV and my secondary 1120mV. Go figure. Over clocking - both cards seem stable at core 940 MHz and memory 1350 MHz. I’m pretty sure there is room for more…

Thinking back my prior system always had some minor stability issues that I always attributed to drivers, software, and such. I’m hoping that this is it.

Thanks for everyone’s help! Still not liking the difference in voltage and thinking of RMAing the card.

New configuration:

SAM_0605-1024x901.jpg
 
You must log in or register to reply here.
Back
Top